Aleena Farooq Email & Phone Number

Lahore, Pk

United States

Driver Performance Event Detection: • Led a cross-functional team in the design and implementation of a risky driving behavior algorithm, achieving a precision rate exceeding 99.5% and a 100% recall for ground truth data. Utilized Python and C for algorithm development. • Conducted in-depth analysis of large volumes of sensor data (Pyspark, Snowflake, and Redash). Enhanced data quality by leveraging various smoothing techniques, and evaluating the algorithm performance downstream.• Formulated a comprehensive strategy for data annotation and established metrics to effectively benchmark algorithms. • Collaborated with multiple teams to build and rollout an end-to-end pipeline, and effectively communicated the algorithm's logic and impact, setting clear expectations across the organization.Collision Detection: • Led a team to train and deploy an ensemble model for detection of collisions in real-time from telematics data, achieving over 90% recall for high-severity collisions. Earned recognition by winning a company-wide Hackathon with the developed model. (Scikit Learn, H2o) • Initiated a data collection initiative to proactively detect data drift, facilitating model retraining and optimization. • Spearheaded a cross-team effort to build a pipeline, optimizing for reduced response time, resulting in a latency of <8 seconds.

Pakistan

Call Transcription Analytics Engine• Leading a group of 10 people through automating the process of call quality assurance (Python, Pandas, MySQL, Docker)• Producing ~5x coverage of call volume and ~10x efficiency per call evaluation, using the automated ML pipeline• Gathering requirements and identifying gaps, by mediating between the client(Call Centre Agents) and development team• Employing Conversational AI and NLP, using rule based techniques and deep learning models on unstructured text data (Sklearn, TensorFlow, Keras, PyTorch)• Using dialogue state tracking for slot labeling and maintaining belief states across conversations• Working with SOTA pretrained models using transfer learning, for generalizable solutions and better mitigating the noise in audio transcription data (BERT, GPT, T5)

Pakistan

Fraud Detection Engine• Developed and deployed, from scratch, a complete configurable pipeline for financial fraud detection using machine learning and distributed computing tools (Python, Apache Spark, Scala, Hadoop, Docker)• Processed data for ~40 million transactions, achieving ~5x improvement in diagnostic odd ratio• Performed exploratory analysis on Big Data (Time Series) to extract insights for engineering new features (Informix, PostgreSQL, Greenplum, MySQL)• Applied multiple ensemble modeling techniques for supervised learning (Gradient Boosted Trees, Random Forest, LSTM, XGBoost using Pyspark, ML, MLlib)• Incorporated unsupervised learning to detect anomalies in cardholder behavior and identifying spending patterns (Clustering, Histogram Based Outlier Score)• Improved prediction pipeline efficiency by 200%, for real time fraud prediction

Pakistan

• Trained and facilitated students from diverse professional/academic backgrounds to get acquainted with Data Science concepts• Familiarized students with data science starter toolkit (Python, Jupyter Notebooks) along with libraries for data handling and machine learning algorithms, concluding with a ML project for evaluation

Lahore, Punjab, Pakistan

Course: Artificial Intelligence(CS401)• Graded and evaluated a class of up to 50 students• Conducted interactive learning sessions

Webify.io - a DIY, code-free website builder that allows anyone to set up a mobile optimized web presence in under 30 minutesDesigned and integrated a database conforming to the requirements of the project

Bachelor Of Computer Science - Bscs, Computer Science

Activities and Societies: NUCES-ACM, IEEE Students Chapter, Softec

Ics, Mathematics, Physics And Computer Science